Go Back to Clockwork

A Clockwork ARM1G

Two or three years ago I was chatting to a friend about clockwork locomotive power. We have an April get-together named the Spring WindUp, which has competition runs for fastest, strongest, etc. The issue of longest distance came up, and it was decided that what is needed for long distance is a long spring. I started to think about it, and what follows here is what evolved from that first discussion. I doubt very much that the ideas here are new; but I chose not to do what is known as a literature search; instead I decided to (re-)invent, simply for the fun of it.

Go Back to the Top
The Basic Idea
Go Back to the Contents Links
The problem with a long spring is its length: the coils get bigger and bigger and, soon, you hit the cab roof or run out of loading gauge. On the other hand a locomotive, steam or otherwise, can be long, a 2-12-2 is very long. Can the two lengths be brought together in a useful way?

Imagine a long locomotive boiler full of clock springs all mounted on a shaft that runs down the centre of the boiler, so that the spring coils are essentially concentric with the boiler shell. Now, starting at one end, imagine the winding key being plugged into the inside of the first spring, and the outside of that spring connected to the inside of the second spring by another, internal, key. Now repeat that connection for all the springs in the boiler, connecting the second spring to the third, the third to the fourth, and so on; the outside of the last spring being connected to an output shaft.

Imagine turning the first winding key with the output shaft fixed so that it cannot move. The inside of the first spring turns, and the torque is transmitted through the spring to its outside, and, via the internal key, to the second spring, thence to the third, and so on. Thus, as the first key is turned, the torque, and the contraction of the convolutions it causes, propagates down the springs; each spring taking up its share of the contraction.

That is the basic idea. What it comes down to is connecting springs in series. In contrast, if all the insides of the springs were connected to a central shaft, and all the outsides were connected to an outer shell, then the springs would be connected in parallel.

If N springs are connected in parallel, then the output torque is N times the torque of one spring, and the contraction of each of the N springs is the same as the contraction of one spring. If the springs are connected in series, the torque of the N springs is the same as the torque of one spring, and the contraction of each spring is one Nth of the overall contraction. Thus, connecting springs in series, in effect, produces a single long spring.

There is a remaining issue, which is that the output shaft is perpendicular to the driving axles of the locomotive. So, a right-angle drive of some sort is required.

The First Locomotive
Go Back to the Contents Links
The Spring WindUp was a few weeks away, so I built a proof-of-principle engine from the usual collection of bits-that-I-had. The springs were purchased from a clock parts supplier who happens to live a few miles from me. By good fortune, there is a type of spring that is supplied within a casing (Hermle), and this made it fairly easy to lash-up a prototype using two springs in series. It did not work. In my haste, I had omitted the outer shell (pseudo-boiler)! I know, I know, but these things happen. Of course, in the design being considered, this outer shell has the important function that it has, so the reader can imagine the result of its omission.

A year later I dragged out the experimental prototype and remedied the omission. Also I lowered the overall gear ratio, which I was fairly sure I had set too high - recall that I was after distance performance. This time the engine worked, and after a little bending and sloshing oil everywhere, and despite out-of-true and other generally nasty friction generators, performance was there, and the principle was established. This locomotive would run for about 400 feet on one winding, but would not pull anything much. You can see this locomotive run at the Spring Windup 2011.

The images below are of this first prototype. The key spindle and tension-retaining ratchet are on the front; the spring casings can be seen through the holes in the pseudo-boiler casing; the primary ratio-amplifying gearbox is a Marx mechanism that is in pretty bad shape; the secondary amplification cum right-angle drive is clearly seen, these are parts from a photographic slide projector; the chassis is from a steam engine that is waiting its turn for refurbishment. A construction detail is that the drive train is reversed from the description above, i.e., the winding key plugs into the outside of the spring, and the inter-spring internal key connection is from the inside to the outside.

The Second Locomotive
Go Back to the Contents Links
I wanted to build a proper model locomotive, albeit still an experimental prototype, using the springs-in-series principle, but also wanted to cut down on the work not directed at developing this principle. I began to look in the G1MRA Journal for a kit of some kind that would be suitable, and made a couple of enquiries, before deciding on ARM1G as a host. I grew up on the Southern Region of BR, indeed have ridden on and fired an H class as a teenager, so that fitted nicely.

An H class is not ideal for the project in hand for two reasons: first because the boiler is relatively small in diameter, and lots of boiler volume for springs seems desirable. And, second, because the distance from the boiler backhead - where the spring output shaft exits - to the rear driving axle is rather large. This distance necessitates a long geartrain, some gears of which are idlers. But, the advantage of the availability of commerical wheels and other parts was appealing. And so I laid out the General Arrangement of my clockwork version of ARM1G in the winter of 2011/2. This was followed by months of precisely nothing due to other pressures.

The image below shows the general arrangement as it pertains to the present writing. What can be seen are seven springs; winder and ratchet/pawl at the smokebox end; a crown gear and pinion (http://www.zakgear.com/Scrown.html) at the cab end. The crown gear and pinion form the required right-angle drive, as well as providing ratio amplification. The output gear of the right angle drive is the beginning of a long chain of spur gears terminating at the rear coupled wheels axle. The reverser, governor, etc., are not shown.

Making It
Go Back to the Contents Links
There are four main assemblies that make up the second prototype locomotive: the locomotive chassis, which is where ARM1G comes in; the pseudo-boiler containing springs; the transmission, including gear ratio amplification, governer, and reverser; and the H class superstructure and other cosmetics, which makes the machine look like a steam engine built one hundred years ago.

Only the first three assemblies are necessary to make a functioning locomotive, and these are described below. The cosmetic work will follow in due course; but it is deemed not to have the same interest value as those of the other assemblies.

Go Back to the Contents Links
There is not a lot to describe with regard to the chassis. The overall dimensions, wheelbase, etc., are, intentionally, identical to those for ARM1G. However, it turned out that the mainframe details are quite different because of the different requirements. Also, the stretchers were made differently, because it was easier for me, lacking a mill, to make stretchers using hexagon rod, and K&S tubing and washers, and turn the ends in a lathe. The important functional difference lies in the configuration of the driving axle, which has a gear and two supports to which the transmission is attached. The marks in the left support are a mystery; but we do have large mice around here.

This image shows the completed chassis; I used Slaters wheels per the ARM1G specification. The transverse member at the front is a support for the pseudo-boiler.

The Spring Loaded Boiler
Go Back to the Contents Links
A pseudo-boiler loaded with springs is an energy storage, but not an energy generating, device. So, for brevity's sake, and until a better name emerges, the pseudo-boiler will be known as a Spring Energy Device, or Sprend, or possibly Sprendev.

The General Arrangement image above shows the layout of the sprendev. The image below shows a tray containing all the parts for the sprend prior to assembly. Starting from the black outer shell, and working around clockwise, are seen a spring, as purchased; an empty bell for that spring, this is the bell that has the winder ratchet built into it; a row of six spring bells, already loaded with springs; ending with the housing for the ratio amplifying, right-angle drive, that is the power supply end of the sprend. Continuing up the middle of the tray, are seen the crown, final drive, and pinion, gears that are mounted in the housing with the journal bolt. The remaining items are a shim washer for the right-angle drive; the winder support with its large diameter bearing that clears the ratchet periphery, and the associated pawl and its spring; and, next to the springs, the long central shaft that supports and locates the spring bells.

The clearance slot in the end of the outer shell is to miss the pinion. This slot has no other function; it became necessary because of the sizes of the tubes that I obtained to make the outer shell and the spring bells. In an ideal world this slot would not exist.

The image below shows the stack of internal parts of the sprend assembled during development and nominally ready for loading. More work has been completed, including machining the outsides of the bells (see above), to produce the current, working, sprend.

The sprend looks quite innocuous and boring once assembled. However, it is heavy since it is full of steel; and it does store quite a lot of energy. This image shows the wind end of the sprend, including an adjustable pin that bears on the chassis support shown earlier.

Go Back to the Contents Links
The basic shape of the transmission is very ordinary; the image below is a view looking up from the bottom. The axle connection is on the right; the governor is on the left; the wide two-piece gears on long shafts are the sliding mesh reverser; the other gears are idlers, all necessary for the reason given in The Second Locomotive; the connection to the sprend is hidden behind the reverser gears.

The sprend crown gear has 76 teeth and meshes internally with a pinion with 11 teeth; this 11 teeth gear is on a common shaft with the 64 teeth sprend output; which meshes with a 24 teeth gearbox input on a common shaft with a 48 teeth gear. After a string of idlers the final mesh is with a 29 teeth axle gear. Thus the overall amplifying gear ratio is 76/11 * 64/24 * 48/29 = 30.5

The overall ratio is changed by changing the transmission side plates drilled to accept an appropriate set of gears; thus the transmission is completely stripped and re-built. The connections to the sprend and the drive axle are made with a few screws. The gear in the middle of the image meshing with the reverser gear is designated the gearbox output and no gear from that one, back to the sprend, is intended to be changed. Any gear between the gearbox output and the axle, except the axle gear, is changeable to get the required overall ratio. With hindsight, it might have been better to design the transmission in two parts to reflect this concept. The image shows a 30.5 ratio gearset; however, originally there was a 24/48 step in the idler line giving an overall ratio of 61. This re-work avoided the need to make new side plates; which was useful since I mark out and drill by hand; a mill-with-DRO would make this an easy task. Bearing in mind that this is an experimental locomotive, this re-work was deemed to be an acceptable procedure despite the insertion of a small idler pinion. The brass strips on the side plates also are a re-work, caused by a decision to re-site the governor. The long screw that sticks out is a temporary adjustment control for the governor.

Most of the transmission parts are purchased, this includes the bearings which are self-aligning press-in sintered bronze at $1.05 each! The sliding mesh reverser has no actuation mechanism, yet; and the governor is an axial-loading type. Both the governor and the reverser are based on design concepts due to John van Riemsdijk.

Go Back to the Contents Links
The twelfth of December turned out to be an important day. This was when performance was to be checked in what was expected to be a somewhat definitional way. The first track run, a couple of weeks earlier, had established that an overall gear ratio of 61 was too high, at least for the state of the locomotive as it was at that time. With no load, the distance travelled was quite good at over 400 feet, but a good push was required to start, and there was no reserve power. The current track run was with an overall ratio of 30.5. And the hope was that the locomotive would haul properly two or three coaches on a friend's level track (mine has significant gradients), with the conclusion that a useful ratio range was established. Such a conclusion would enable suitable performance targets to be established. The first track run had established that the total available number of wind turns of the sprend is over 60, i.e., about nine turns per spring. I was very hopeful that I could settle on 39.2 as an overall ratio, since, with 2.1 inch diameter wheels, this would require 61.2 sprend turns for a track run of a quarter of a mile (402 metres). I had a target train of at least two coaches in mind: my clear memory is of H class locomotives with a two coach train, three sometimes, and none at all with four coaches.
A Forced Stop
Go Back to the Contents Links
Winding sixty turns against a fairly powerful spring is not a job to do by hand. From an early date I had decided that a power drill with a hexagon socket was the way to do the winding. I thought that a torque limiting chuck on the drill would be an effective way to limit each wind. It turns out that this is not a good way to control a wind, although it works. The problem lies in the hammering caused by the torque limiter. This hammering is difficult to control for the human holding the locomotive, and probably is bad in the longer term for the relatively soft ends of the springs. On the day of the definitional performance runs, the hammering broke a soldered joint in the winder, so that the ratchet and pawl no longer worked; and this ended the tests and caused doubts to be thrown on my soldering abilities.

Unfortunately this failure occurred in the context of my moving house next spring and the need to get ready for that move. This means that the whole project will be delayed for many months by the breakage. But I decided to continue writing these notes to record the present state of affairs, both for when I restart, and in case someone else would like to try The Basic Idea.

Go Back to the Contents Links
A little running was carried out before the breakage; the results were encouraging but not definitive. The image below shows a train that ran for over 500 feet; a small nudge was required to get the train started. Removing the coach at the end of the run allowed the locomotive to run for another 300 feet or so. The overall ratio of 30.5 gives about 50 sprend turns; and this may be right, analytical accuracy was not the concern at this time, and, probably, the sprend was not fully wound. A two coach train required just a little too much from the locomotive and the train stalled twice on each of the 143 feet laps, on the curves; but it would accelerate on the straights.

Go Back to the Contents Links
As indicated, this project has been put on the shelf just as it was getting interesting. I plan on returning when life allows.

The Basic Idea is sound. The problems that are reported here are with other aspects of the second prototype: the chassis, the transmission, and implementation details of the sprend. The possible improvements that are postulated here are about implementation design changes. I have little doubt that the quarter mile, two coach, target can be achieved with the ARM1G design, and rather expect that considerably more is achievable with a large boiler prototype.

It is intriguing to consider that a quarter of a mile completed at a "scale" 60mph is an eight minute run.

What follows is a few notes concerning pressing development issues; more could be written. I suppose this is where I am re-visiting problems that were solved a long time ago. These notes will be useful when I restart, and may be of interest to others - who may know the answers.

  • A major puzzle is the lack of range in the light of a lot of residual wind when the train stops. It appears that there is a lot of resistance due to friction, and the question is what is the avoidable/unavoidable ratio. It is clear that a small amount of resistance, judged on a human scale, has a large effect. A simple example lies in the governor, which governs by using friction, and yet does not get hot.
    • Old commercial clockwork mechanisms tend to have tiny pin bearings; and this leads to thinking about wind-up watches and jewelled bearings. The subject locomotive has about twenty 1/8 inch journal bearings in the transmission, and these normally run smoothly; but they may be providing too much drag. The solutions that could be tried include:
      • Journals machined down to tiny, per the old commercial mechanisms
      • Needle roller bearings! These are intriguing, being smaller than the sintered bronze used to date. However, they cost $9 each instead of $1. Hmmmm.
    • The transmission side frames flex, and the slight, not visible, movement increases the resistance; perhaps by loading the journal bearings. This effect, or perhaps a different, but similar, one, can be realized by holding the locomotive upside down and rotating it - on its side, vertically, etc.; the effect is to change the running speed of the mechanism, sometimes to the point of stopping it. The frame flexing effect also can be demonstrated by flexing the frame with one's fingers. A design change could be to improve the support provided by the side frame stretchers in order to stiffen the structure. In any case, the running speed of the mechanism should not alter depending on how the locomotive is positioned in space; clearly this is a problem to solve. It is possible that this effect is the one causing the locomotive to slow down on curves and speed up on straights.
    • Sprend internal jolting sometimes occurs during unwinding, this suggests that the sprend spring coils are sticking to some degree and not sliding nicely. It is not clear that this matters much, except aesthetically. Using different lubricants may help.
    • It is very often - not always - necessary to nudge the mechanism to get it started, on the track or on the workbench; thereafter the mechanism will accelerate, often to a high speed. In other words, the resistance appears to go down with increased speed.
  • The wind-up mechanism needs to be addressed. Current thinking is to have lugs on the wind end of the sprend and a mating housing on the power drill; this arrangement will route all the winding forces through the metal and eliminate the current human component. A turn counter is desirable, as is elimination of the break-style torque limiter.
  • The centre of gravity of the locomotive is almost exactly over the driving axle. Since the design has no axle springing this means that the coupled axle loading is controlled by the bogie springing. From the paragraph about bogie springing in the ARM1G manual, it appears that this is true of the steam ARM1G as well. It may be worthwhile to add weight at the front of the locomotive.

Of course, I have started to think about a third locomotive; and have sketched out a General Arrangement. It is a freelance Baltic tank with a very large sprend and a shorter transmission than the current one. I think that I have found a way to eliminate the spring bells (a description follows in A Speculative Idea), which will enable more room for spring expansion, and, perhaps cut down on sprend internal friction. Also, the right angle drive can be a normal bevel drive.

As for potential performance it is interesting to compare spring specifications.

  • The first prototype used two springs each 21mm x 0.42mm x 1.90m x 47mm diameter, and yielded 400 feet running light.
  • The second, current ARM1G, prototype uses seven springs each specified as 20mm x 0.40mm x 1.35m x 38mm diameter. These springs are smaller than those in the first prototype, but there are more of them. This locomotive yields 500 feet loaded, with the expectation of at least doubling that distance.
  • The third, speculative, prototype would use sixteen springs each specified as 16mm x 0.60mm x 2.74m x 60mm diameter.

A Speculative Idea
Go Back to the Contents Links
The implementation of The Basic Idea in the Clockwork ARM1G is somewhat clunky, and perhaps, although unwittingly, was overly influenced by the Hermle clock enclosed spring design. What follows is a different configuration concept that may be better suited to locomotive sprend application.

The springs are arranged in pairs; these springs are arranged to coil in opposite directions. By connecting the springs at the outer ends, winding to tighten the coils of the first spring also tightens the coils of the second spring.

One pair is one spring cycle, and the cycle is repeated as many times as desired. Input and output are at the inner ends of the spring-pair, which enables simple coupling between pairs.

There is a practical consideration caused by the axial relative displacement, which causes a transverse twist where the springs are connected. Probably this twist will need to be resisted by a bearing on the central shaft, rather than the conceptual rod-and-loop connection shown in the sketch. A balancing, second kind of, pairing is not obvious.

Go Back to the Contents Links
Despite the external pressures preventing full attention to this project, a little progress was made in the early part of 2013. The main driver and objective of the further effort being not to leave the locomotive in a broken condition.
Force Circuit Winder
Go Back to the Contents Links
Since it was the cause of the breakage, the problem with power winding described in A Forced Stop has been addressed and a newly designed winder has been made. This is referred to as a Force Circuit Winder (FCW) because opposition of the electrically generated opposing torques at the chuck and the power drill casing is resolved by a casing extension that connects to the locomotive. Thus the basic physics requirement of a force circuit is met by replacing the human muscle and bones connection, normal when using a power drill, by a metal tube and fasteners. The unwieldy characteristics of the original system thus disappear completely from the human's point of view, including the need to hold onto the locomotive.

The further important feature of the Force Circuit Winder is the incorporation of a revolution, or turn, counter. Thus, setting any specific wind, in particular limiting the maximum wind, can be accurately controlled as a number of turns instead of relying on the power drill torque-limiter.

The first image below shows the winder parts, complete but unpainted. The three shafts are commercial items as used in the construction trade and elsewhere. One of these shafts has been fitted with an eccentric. This shaft runs in the bearing in the end of the main casing, and the eccentric drives the turn counter. The second image shows the painted and assembled winder; and the third image shows the three pin bayonet connector system used to transmit torque between the locomotive and the main winder casing.

Go Back to the Contents Links
Repair to the locomotive was limited to re-soldering the ratchet wheel; replacing the first spring in the sprend, which had indeed started to tear; and re-working the pawl engagement springing. New work was fitting the three shoulder bolts which serve as the locomotive side of the torque transmitting bayonet connector.

It was necessary to re-work the pawl engagement springing in order to allow clearance for the winder side of the bayonet connector. This was a little difficult because of the lack of room and the realities of a retro-fit. The result seems to be acceptable as part of a prototype, and installation is not expected to be difficult on a future design.

Performance II
Go Back to the Contents Links
The new winder was successful and useful immediately; its trials are just beginning, but it has had a good start. One immediate result was enabling accurate measurement of the maximum number of winds that can be input to the sprend. During the first runs of the repaired locomotive it became clear that the maximum number of turns that can be input to the sprend is 54. Consistantly, on several occasions, attempting to input one more turn would cause the torque-limiter (set to its maximum torque) to break; the counter would increment, but it is unlikely that the springs tightened any more.

It happened that I was a participant at a train-show immediately after doing the repairs described above. So I took the opportunity to do some testing on the essentially level track during the two day event. Below are my notes made just after those tests.

Given the measured TCA track length of 87 feet, yesterday the ARM1G ran 700 feet with two coaches; that clears the 1/8 mile mark. Today it ran, several times, for over six laps with three coaches, and then a further three-plus laps with one coach; it cleared nine laps easily a few times, but not quite ten laps. However, it will do ten laps (880 feet) running light. The problem is that the locomotive is too fast now, hence the three coach load. I was getting 54 sprend turns pretty consistently; and that ties up very well with the distance. With 2.1 inch wheels and a 30.5 overall ratio, 54 turns yields 905 feet. So, I shall not achieve the quarter-mile, two coach, target, but that did not have much basis anyway. However, it looks as though 3/16 mile (990 feet), or the magic 1000 feet, with two coaches, is possible, as is 1/8 mile with three coaches.

54 sprend turns for a 7 spring sprend is just shy of 8 turns per spring. This contradicts the earlier guesstimate of over 60 turns, i.e., about 9 turns per spring. And this, in turn, makes the target locomotive performance of a quarter of a mile look to be out of reach.

At the train-show there was an improvement in performance as time wore on, suggesting that something was settling-in. I speculate that this is not surfaces smoothing out, as might be a first guess, but rather movement of the gearbox frames and the self-aligning bronze bearings. It has been demonstrated that the mechanism can bind somewhat with finger pressure on a side.

Whatever the cause of the loosening-up, it is becoming a problem because the governor is so ineffective; excessive speed at the beginning of a run being the result. It will be necessary at some point to re-site the governor in order to run it at a higher speed. At present the governor runs too slowly for its centrifugal force to change sufficiently with locomotive speed.

Here is the locomotive running at the train-show.

Performance III
Go Back to the Contents Links
Oh dear, it broke again. A few days after the train-show I went to my friend's track again; the plan was to produce a video. Interestingly, performance was not as good as it had been at the train-show; but it was far from bad: we did see at least 500 feet with two coaches, I think it was considerably more. Frankly, I forget, I am writing some time after the event. We did get some video, although now the focus was dealing with lighting and other photographic production problems; and then the camera battery died. While the battery was charging, I played around with the train; and then I, with the FCW, broke something in the locomotive.

Two issues can be addressed at this point. First is the maximum number of sprend turns. Based on the experiences of the torque limiter bashing-up the springs and the physical evidence of a torn spring, I believe that 54 sprend turns puts a very considerable load on the attachments at the ends of the springs, particularly the spring closest to the winder. Thus, I have decided, for the future, to limit the number of sprend turns to 50-51, which is just over seven turns per spring. Given the overall amplifying ratio of 30.5 this gives a theoretical range of 838-854 feet.

The second issue is the current lack of an effective governor, resulting in excessive speed at the beginning of a run. In response, I have formalized a new operational procedure to be used until the governor is more effective. The plan is to start off with a relatively heavy load, thus governing the speed, and remove coaches on the fly as the run progresses. I plan to keep to my two-coach target. Thus, for recording hauling performance, whatever speed controlling load I start off using will be reduced only down to two coaches, running to a complete stop. In addition, I shall keep a record of light running distance; this will be accomplished by further reducing the two coach load down to zero and a complete stop, and using whatever means is necessary to control speed.

With regard to excessive speed, I note that, whilst the camera battery was charging, on one occasion I saw the locomotive lift its inside wheels. I saw it head-on, and I have a very clear memory of the moment. Fortunately, the locomotive reached a straight piece of track just in time.

With regard to the failure. It turned out that the soldered joint at the winder had broken again, although in a different place. The joint is right next to the winder bearing and it appears that I really held back with solder application to avoid getting solder on the bearing surface; there was very little solder in the joint. So, that has been repaired, and when the snow melts we can have another go at a video. It is clear, however, that there are significant loads on the winder end of the sprend.

Contra-Spiralled Spring
Go Back to the Contents Links
A test rig has been created to perform proof-of-principle tests on A Speculative Idea. The subject springs now are known as Contra-Spiralled Springs (CSSs). These notes are to record the results of the first test.

The test spring wound as expected with two difficulties. One difficulty was that the coils did not stay concentric, but fell sideways into a ball, rather like a ball of knitting wool. It seems that this problem can be resolved fairly easily with a suitably designed barrier between every pair of spring coils. These barriers can be very thin material, probably metal, and perhaps just 0.005-0.010 inches thick, because the sideways force is very small. In the images below the plastic discs, brown and white, performed the required function, albeit not very well because the discs used tended to split around their centre holes. The blue tape also is performing the same function, but, of course, is static.

The second difficulty is more serious. The method of connecting the outsides of the two component springs to form a single CSS is inadequate because it is too rigid. The function of transmitting the tension between the two springs whilst resisting the generated transverse torque appears to be handled well; although the loads applied during the test were low. However, the separate plate method used to join the component springs produces a stiff joint in the, otherwise supple, spring. This stiffness causes the offsetting of the centre axes of the coils shown in the images. Also, it is fairly clear that high stress in the springs, at the ends of the plates, will develop as the spring is loaded further.

The joint needs to be re-thought. A one-piece spring probably would be the very best arrangement; but has obvious production difficulties. Possibly, if the transverse twist loads are small enough, the schematic concept using loops and a pin would work. Spot-welding thin spring material is another idea.

Overall, the results from the first test are encouraging. The spring behaved as expected with respect to its main function, and the difficulties encountered appear to be surmountable. Usable torque was generated in the main shaft; however, the loads were light. It is not known yet if practical loads can be generated without serious problem.

Performance IV
Go Back to the Contents Links
We had another go a producing a video with the repaired locomotive, and this time we were successful.

The video is largely self-explanatory and the initial shots are simply moving versions of the images contained above. However, it is worth pointing out that the Force Circuit Winder (FCW) continues to work well, as is seen. It is a little awkward to engage the bayonet connection without touching the locomotive, which will be undesirable when there are cosmetic details and finish paint added to the locomotive. However, the awkwardness is due mainly to the winder being quite heavy and considerably heavier than the locomotive. Also, the power drill in use has a speed jump as its speed is increased, clearly evident; but this is nothing to do with the design of the FCW itself, nor the locomotive, which is on the receiving end of everything.

  • The Continuous Run
    Go Back to the Contents Links
    A little commentary is appropriate for better understanding of the continuous run in the video. The first thing to note is that the sprend was wound with 50-51 turns, this cannot be seen. Next, is that the two coach train is pushed to get it to start. Very occasionally, the train is self-starting. The problem is thought to be stiction, a well-known railway difficulty; and it may be possible to alleviate it in this case with a different spring lubricant, such as graphite.

    Once started the train accelerates over the straight metal bridge, and then - and this is not obvious in the video - it slows slightly as it goes round the first, and every subsequent, curve. This slowing is thought to be caused by a mild distortion of the gearbox structure, caused by the track curvature, which binds the motion somewhat. So the progress around the track is a series of accelerations and decelerations, dictated by the shape of the rails! This go-slow-go-slow motion will result in the train stalling just before two laps are completed, so you will see a deft crane-shunt early enough to prevent a stop; this leaves the locomotive with just one coach to pull.

    Now the locomotive, with no speed governor, is under-loaded; and the one coach train gets up to a fast, very unrealistic, mildly worrying, speed for well over two laps. But, in a similar fashion to three laps earlier, the train will stall just before five laps with one coach, so another crane-shunt is effected, leaving the locomotive with no train. One more lap ensues and the locomotive stops somewhat short of the six lap mark.

    The run in the video actually is on the short side, and one or two runs that day went over six laps. But it is a record of a typical run, and the cameraman was getting dizzy.

  • Video
    Click on the image to run the video in another window, leaving this one for reference.

  • Some Numbers
    Go Back to the Contents Links
    • The track circuit is about 143 feet long
    • The locomotive weighs 5lbs
    • Each coach weighs 4lbs
    • The total distance travelled by the, ungoverned, locomotive was nearly six laps, about 830 feet. The time taken (from the video) was 182 seconds. Thus, the average speed was 4.56 feet per second, or 3.11 mph, which many would consider to be a "scale" 99.5 mph. The target speed for this train is 50-60 mph, which is more appropriate for a South Eastern and Chatham Railway, Wainwright H class tank engine pulling a two-coach train between Tonbridge and Paddock Wood.
    • The fastest lap, the second with a one coach train, was completed in about 22 seconds, which is an average speed of 4.4mph, or a "scale" speed of 142mph.

  • The Next Changes
    Go Back to the Contents Links
    There are two changes that stand out as being required, and one change that will explore new ground.

    The obvious required change is the provision of an effective speed governor. The belief is that the current design is good; but the governor is poorly placed in the gear train; it is close to the sprend end, and it should be at the drive end of the gearbox. The governor does not spin quickly enough, and so the change in centrifugal force with speed change is not adequate to provide effective braking load and, thus, effective speed control.

    The less-obvious change is stiffening the gearbox assembly. It still is very easy to stall the locomotive by light finger pressure on the side of either of the gearbox frames. The hope is that the current train slowing on curves will be reduced by gearbox stiffening; and that this,in turn, will enable a much longer run with a two-coach train.

    Both of the above changes can be effected by relatively minor design changes to the gearbox, and manufacture of new parts.

    New ground will be explored by fitting a spring set made from Contra-Spiralled Springs. This new spring set has been designed to be a simple retro-fit to the sprend in the existing locomotive. If these springs work as intended, implementation of the The Basic Idea will be considerably simpler. Also, there will be more room in the sprend for spring expansion due to the elimination of the Hermle-style bells.

Go Back to the Contents Links
I have received considerable assistance in this endeavour from a friend who lives not far from me. And this acknowledgment is made with thanks for use of his track, gear cutting, turning of items too large for my lathe, and for his part as sounding board, critic, and source of information. Marc also is the video producer.

Marc is the editor of Garden Railways Magazine, founding it over twenty-five years ago. And Marc has his own Sidestreet Bannerworks website.

Go Back to the Contents Links
I am writing at the end of January 2015 - a little short of two years since I stopped work on this project. I have moved house, and have built a new workshop, which enables me to restart workshop activities. However, I have decided to change direction somewhat on my clockwork endeavours. This means that development work on the second prototype, the ARM1G, is terminated. I am not abandoning the locomotive; I have done it the courtesy of making it respectable: I have fitted a governor, and completed the reverser; and I have fitted a superstructure to make it model an SECR H class locomotive. But trying to make it go better has stopped.

The reason for stopping has come out of mulling over the problems to be solved. What it all boils down to is that solving all the this-and-that problems, large and small, means doing a lot of re-design and making a lot of new hardware; and I think that it makes more sense to put that effort into designing and building a new locomotive. This new work is described in another section parallel to this one.

The more important development issues that have led to this decision to change direction are mentioned below. There are many other items that are not mentioned below, but which arise in the new design; for example, changing the engagement system between winder and sprend input shaft from a hexagon drive to a self-engaging pin-and-slot.

  • The new governor works, but is too small and not adequately controllable; moving it requires significant re-design and manufacture.
  • I have decided that the location of the extra friction mentioned more than once, explicitly and implicitly, in the text above, e.g., in The Continuous Run, is at the connection between the gearbox assembly and the driven axle. The problem is poor design, rather than the gearbox flexing within itself or something like that, which is what I have been chasing. The problem is that the gearbox and the locomotive mainframe both have bearings on the driven axle, for the purpose of aligning the final gear engagement. Independently, either the gearbox or the frame allow the axle to run freely, but any mis-alignment, static or dynamic, immediately causes binding and extra resistance. Four bearings inline on a stiff shaft is an engineering "known problem" that I should have seen earlier. Repairing this problem requires significant re-design and manufacture.

So the Clockwork ARM1G now becomes a very successful prototype that has reached its retirement point. I hope to run this locomotive in the future as acknowledgment of its contribution to a better-yet clockwork locomotive design.

Cosmetic Completion
Go Back to the Contents Links
This is an on-going log of the progress of completing the locomotive, largely in pictures. I have recorded only the things that I think might be of interest.

The superstructure in the images is made from nickel-silver etches produced by Orion Models, see SECR H-class. The etched superstructure was modified a little to fit what I had produced since the intention of Orion Models was to produce a steam powered locomotive, i.e., a true ARM1G, and my frame configuration is slightly different.

This image shows my front buffer beam with coupler from Kadee and buffers from JustTheTicket. I did not like the original conical buffer contour, on the right, so I turned the buffers, sample on the left, to look more like those on this locomotive. Notice that the rear buffers in that image are larger in diameter than the front buffers. Sometimes the front buffers were larger, too.

The coupler is American; I like Kadee couplers. These couplers are not prototypical, of course - but there are other considerations when actually running gauge 1 models on a garden railway.

Click, or [Enter], on this, or any, image that follows to see a higher resolution version. Return to this description with your browser back-button or [Alt+left-arrow].

This image shows the start of the superstructure installation. In particular, it shows the successful mating of the Orion Model etched parts with my clockwork locomotive frame.

The locomotive proper, i.e., not the superstructure, has to be taken apart once more for a little more metal work and some painting.

This image shows progress of the superstructure installation. This is the first time I have assembled etchings; it seems to be working out, although there have been scares along the way. As shown, the locomotive balance point is close to the rear driver axle, there is just about no weight on the front axle. I think that I shall need to make a heavy chimney.

Another Restart
Go Back to the Contents Links

I am writing in May 2024, about fourteen years since this project was started. Interruptions include building a portable railway. I intend writing as though this project was unaffected; but I expect there will be discontinuities for the reader.

Cosmetic work has been accomplished from time to time; and the images below show the results. I have never painted lining before, as can be seen from the result. Also, the sprend is broken, which troubles me; I do not like to leave the locomotive in a disabled condition, even though I do not plan to run it very much.

Contact Information
Go Back to the Contents Links
Go Back to Clockwork

last-modification-date: 17 May 2024